real time transcription - gradio audiostreaming feature ready

[sandorkonya]

@fedirz thank you for this repo, so far the best one i found for live transcription, only one of the few that utilizes localagreement!

1. The gradient demo on Huggingface features recording audio, but an upcoming gradio merge is going to allow real time transcription directly from browser!

2. could you add flozi00/distilwhisper-german-canary to the available models?

regards

[fedirz]

1. Thanks, I'll take a look once 5.0 is merged in
2. This project supports only models which can be run with ctranslate2. flozi00/distilwhisper-german-canary isn't compatible with ctranslate2 as far as I can tell. The author does have a model flozi00/whisper-large-v3-german-ct2 which looks like it's ctranslate2 compatible, however the model doesn't have proper tags and that's why you don't see it in the list of models when running curl -s http://localhost:8000/v1/models | jq '.data[].id'. Consider asking the author of the model to add a ctranslate2 tag to his model.
   For reference, here's a piece of code which determines available models.

models = huggingface_hub.list_models(library="ctranslate2", tags="automatic-speech-recognition", cardData=True)